arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.13172cs.LG
收藏

具有延迟Bandit反馈的对抗MDP的近优后悔

Near-Optimal Regret for Adversarial MDP with Delayed Bandit Feedback

University of Southern California(南加州大学);Tel Aviv University(特拉维夫大学);Google Research(谷歌研究院);Amazon(亚马逊)

Tiancheng Jin, Tal Lancewicki, Haipeng Luo, Yishay Mansour, Aviv Rosenberg

摘要:

强化学习(RL)的标准假设是智能体立即观察其行为的反馈。然而,在实践中,反馈往往是延迟观察。研究了具有未知转移、逆向变化代价和无限制延迟反馈的偶发马尔可夫决策过程(MDP)的在线学习。更确切地说,针对情节$k$中的代理的反馈仅在情节$k + d^k$结束时被揭示,其中延迟$d^k$可以随情节而改变并且由不经意的对手选择。我们提出了第一个算法,实现接近最佳的$\sqrt{K + D}$遗憾,其中$K$是情节的数量和$D = \sum_{k=1}^K d^k$是总延迟,显着改善后,最有名的遗憾界的$(K + D)^{2/3}$。

英文摘要:

The standard assumption in reinforcement learning (RL) is that agents observe feedback for their actions immediately. However, in practice feedback is often observed in delay. This paper studies online learning in episodic Markov decision process (MDP) with unknown transitions, adversarially changing costs, and unrestricted delayed bandit feedback. More precisely, the feedback for the agent in episode $k$ is revealed only in the end of episode $k + d^k$, where the delay $d^k$ can be changing over episodes and chosen by an oblivious adversary. We present the first algorithms that achieve near-optimal $\sqrt{K + D}$ regret, where $K$ is the number of episodes and $D = \sum_{k=1}^K d^k$ is the total delay, significantly improving upon the best known regret bound of $(K + D)^{2/3}$.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑